We present an efficient snapshot mechanism, briefed as Live Save, to make real time backup of the virtual machine state to a local host. Live Save will iteratively send the state data, store the snapshot file in a local host and, when necessary, send the entire file directly to a remote host – to reduce network bandwidth consumption in existing snapshot practices. In iterative transmissions, when dirty frequency rises to a certain high level, iterations may fail to reduce dirty pages and hence become redundant. To handle the problem, we add a simple judging measure to Live Save to form an Improved Live Save which will adaptively adjust the maximum number of iterations. By maintaining the maximum number of iterations, Improved Live Save can avoid futile iterations – saving unnecessary iterative computation in Live Save. Extended simulation runs are conducted to evaluate the performance of various snapshot mechanisms (Xen Save, Xen Live Migration, Live Migration Save (VNsnap), Live Save and Improved Live Save) in terms of downtime, total execution time and total pages sent. The results show that, in performing the snapshot practice, both Live Save and Improved Live Save are more resource-conserving than the other mechanisms.
Relation:
Journal of Information Science and Engineering 36(2), p.423-440
